Explore Prompts

Design safety culture practices: blameless reporting, review norms, escalation comfort, and training. Include weekly rituals and how to measure culture health.

Create an ethical review process for high-impact use: fairness, accessibility, harm minimization, and user consent. Include a checklist and escalation rules.

Design containment by limiting action scope: allowlisted domains, read-only modes, rate limits, and staged privileges. Include how to measure whether containment is effective.

Create a quarterly roadmap for recursive improvement: initiatives, milestones, safety deliverables, and staffing. Include an ‘if risk increases’ contingency branch.

Design a feedback system that’s hard to game: structured feedback types, sampling, and weighting. Include how to prevent brigading and ensure minority failure modes are captured.

Create a production readiness review checklist: evaluation results, red-team status, privacy review, rollback tested, monitoring live, and owner sign-offs. Output a go/no-go rubric.

Define a safe logging policy: what to log for debugging, what to redact, retention windows, access controls, and anonymization. Include examples and anti-examples.

Design a zero-trust approach: secrets management, short-lived credentials, environment segregation, and auditing. Include how to test that the AI cannot access restricted secrets.

Write explicit requirements to prevent deceptive behavior: transparency rules, auditability, and logging. Include tests to detect hidden policy violations or manipulation attempts.

Design a safe self-reflection procedure: bounded critique, avoiding infinite loops, and ensuring critique is grounded in evidence. Provide stop rules and performance safeguards.

Create a cross-validation pattern: primary model proposes, secondary model critiques/verifies, and a policy gate decides. Include how to handle disagreements and measure error reduction.

Design a multi-agent architecture with safety: role separation, constrained tools, message filtering, and cross-checks. Include failure modes and an evaluation plan.